This suite exercises a Cloud Foundry deployment using the cf CLI and curl. It is scoped to testing user-facing, end-to-end features.

For example, one test pushes an app with cf push, hits an endpoint on the app with curl that causes it to crash, and asserts that we see a crash event in cf events.

Tests that won't be introduced here include things like basic CRUD of an object in the Cloud Controller. Such tests belong with the component they relate to.

These tests are not intended for use against production systems. They're meant for acceptance environments used by people developing Cloud Foundry's releases. While these tests attempt to clean up after themselves, there's no guarantee that they won't change the state of your system in an undesirable way. For lightweight system tests that are safe to run against a production environment, please use the CF Smoke Tests.

NOTE: Because we want to parallelize execution, tests should be written to be executable independently. Tests should not depend on state in other tests, and should not modify the CF state in such a way as to impact other tests.

1. Test Setup
   1. Install Required Dependencies
   2. Test Configuration
2. Test Execution
3. Explanation of Test Groups
4. Contributing

• Install golang >= 1.11. Set up your golang development environment, per golang.org.

• Install the cf CLI. Make sure that it is accessible in your $PATH.

• Install the log-cache plugin with:

  cf install-plugin -r CF-Community "log-cache"

• Install curl

• Check out a copy of cf-acceptance-tests. It uses Go modules, so there is no need to put it in $GOPATH.

• Ensure all submodules are checked out to the correct SHA. The easiest way to do this is by running:

  ./bin/update_submodules

• Install a running Cloud Foundry deployment to run these acceptance tests against. For example, bosh-lite.

All go dependencies required by CATs are vendored in the vendor directory.

Make sure to have Golang 1.11+

In order to update a current dependency to a specific version, do the following:

cd cf-acceptance-tests
source .envrc
go get <import_path>@<revision_number>
go mod vendor

If you'd like to add a new dependency just run:

go mod tidy
go mod vendor

You must set the environment variable $CONFIG which points to a JSON file that contains several pieces of data that will be used to configure the acceptance tests, e.g. telling the tests how to target your running Cloud Foundry deployment and what tests to run.

The following can be pasted into a terminal and will set up a sufficient $CONFIG to run the core test suites against a BOSH-Lite deployment of CF.

See example-cats-config.json for reference.

Only the following test groups are run by default:

include_apps
include_detect
include_routing
include_v3

• api: Cloud Controller API endpoint.
• admin_user: Name of a user in your CF instance with admin credentials. This admin user must have the doppler.firehose scope.
• admin_password: Password of the admin user above.
• apps_domain: A shared domain that tests can use to create subdomains that will route to applications also created in the tests.
• skip_ssl_validation: Set to true if using an invalid (e.g. self-signed) cert for traffic routed to your CF instance; this is generally always true for BOSH-Lite deployments of CF.

include_* parameters are used to specify whether to skip tests based on how a deployment is configured.

• include_apps: Flag to include the apps test group.

• include_container_networking: Flag to include tests related to container networking. include_security_groups must also be set for tests to run. See below

• credhub_mode: Valid values are assisted or non-assisted. See below.

• credhub_location: Location of CredHub instance; default is https://credhub.service.cf.internal:8844

• credhub_client: UAA client credential for Service Broker write access to CredHub (required for CredHub tests); default is credhub_admin_client.

• credhub_secret: UAA client secret for Service Broker write access to CredHub (required for CredHub tests).

• include_deployments: Flag to include tests for the cloud controller rolling deployments. V3 must also be enabled.

• include_detect: Flag to include tests in the detect group.

• include_docker: Flag to include tests related to running Docker apps on Diego. Diego must be deployed and the CC API docker_diego feature flag must be enabled for these tests to pass.

• include_internet_dependent: Flag to include tests that require the deployment to have internet access.

• include_internetless: Flag to include tests that require the deployment to not have internet access.

• include_isolation_segments: Flag to include isolation segment tests.

• include_private_docker_registry: Flag to run tests that rely on a private docker image. See below.

• include_route_services: Flag to include the route services tests. Diego must be deployed for these tests to pass.

• include_routing: Flag to include the routing tests.

• include_routing_isolation_segments: Flag to include routing isolation segments tests. See below. Cannot be run together with logging isolation segments tests.

• include_security_groups: Flag to include tests for security groups. See below

• include_services: Flag to include test for the services API.

• include_service_instance_sharing: Flag to include tests for service instance sharing between spaces. include_services must be set for these tests to run. The service_instance_sharing feature flag must also be enabled for these tests to pass.

• include_ssh: Flag to include tests for Diego container ssh feature.

• include_sso: Flag to include the services tests that integrate with Single Sign On. include_services must also be set for tests to run.

• include_tasks: Flag to include the v3 task tests. include_v3 must also be set for tests to run. The CC API task_creation feature flag must be enabled for these tests to pass.

• include_tcp_routing: Flag to include the TCP Routing tests. These tests are equivalent to the TCP Routing tests from the Routing Acceptance Tests.

• include_v3: Flag to include tests for the v3 API.

• include_zipkin: Flag to include tests for Zipkin tracing. include_routing must also be set for tests to run. CF must be deployed with router.tracing.enable_zipkin set for tests to pass.

• use_http: Set to true if you would like CF Acceptance Tests to use HTTP when making api and application requests. (default is HTTPS)

• use_existing_organization: Set to true when you need to specify an existing organization to use rather than creating a new organization.

• existing_organization: Name of the existing organization to use.

• use_existing_user: The admin user configured above will normally be used to create a temporary user (with lesser permissions) to perform actions (such as push applications) during tests, and then delete said user after the tests have run; set this to true if you want to use an existing user, configured via the following properties.

• keep_user_at_suite_end: If using an existing user (see above), set this to true unless you are okay having your existing user being deleted at the end. You can also set this to true when not using an existing user if you want to leave the temporary user around for debugging purposes after the test teardown.

• existing_user: Name of the existing user to use.

• existing_user_password: Password for the existing user to use.

• artifacts_directory: If set, cf CLI trace output from test runs will be captured in files and placed in this directory. See below for more.

• default_timeout: Default time (in seconds) to wait for polling assertions that wait for asynchronous results.

• cf_push_timeout: Default time (in seconds) to wait for cf push commands to succeed.

• long_curl_timeout: Default time (in seconds) to wait for assertions that curl slow endpoints of test applications.

• broker_start_timeout (only relevant for services test group): Time (in seconds) to wait for service broker test app to start.

• async_service_operation_timeout (only relevant for the services test group): Time (in seconds) to wait for an asynchronous service operation to complete.

• test_password: Used to set the password for the test user. This may be needed if your CF installation has password policies.

• timeout_scale: Used primarily to scale default timeouts for test setup and teardown actions (e.g. creating an org) as opposed to main test actions (e.g. pushing an app).

• isolation_segment_name: Name of the isolation segment to use for the isolation segments test.

• isolation_segment_domain: Domain that will route to the isolated router in the isolation segments and routing isolation segments tests. See below

• private_docker_registry_image: Name of the private docker image to use when testing private docker registries. See below

• private_docker_registry_username: Username to access the private docker repository. See below

• private_docker_registry_password: Password to access the private docker repository. See below

• unallocated_ip_for_security_group: An unused IP address in the private network used by CF. Defaults to 10.0.244.255. See below

• require_proxied_app_traffic: Set this to true if Diego was configured to require proxied port mappings, i.e. if containers.proxy.enable_unproxied_port_mappings is set to false. Note that this also requires using the cf-syslog-skip-cert-verify.

• staticfile_buildpack_name See below

• java_buildpack_name See below

• ruby_buildpack_name See below

• nodejs_buildpack_name See below

• go_buildpack_name See below

• python_buildpack_name See below

• php_buildpack_name See below

• binary_buildpack_name See below

• include_windows: Flag to include the tests that run against Windows cells.

• use_windows_test_task: Flag to include the tasks tests on Windows cells. Default is false.

• use_windows_context_path: Flag to include the Windows context path routing tests. Default is false.

• windows_stack: Windows stack to run tests against. Must be windows.

• include_service_discovery: Flag to include test for the service discovery. These tests use apps.internal domain, which is the default in cf-networking-release. The internal domain is currently not configurable.

• stacks: An array of stacks to test against. Currently only cflinuxfs3 stack is supported. Default is [cflinuxfs3].

• include_volume_services: Flag to include the tests for volume services. The following requirements must be met to run this suite: tcp-routing must be deployed.

• volume_service_name: The name of the volume service provided by the volume service broker.

• volume_service_plan_name: The name of the plan of the service provided by the volume service broker.

• volume_service_create_config: The JSON configuration that is used when volume service is created.

• infrastructure: The name of the infrastructure for the environment that the tests will run against. Must be either "vms" or "kubernetes". Defaults to "vms".

Many tests specify a buildpack when pushing an app, so that on diego the app staging process completes in less time. The default names for the buildpacks are as follows; if you have buildpacks with different names, you can override them by setting different names:

• staticfile_buildpack_name: staticfile_buildpack
• java_buildpack_name: java_buildpack
• ruby_buildpack_name: ruby_buildpack
• nodejs_buildpack_name: nodejs_buildpack
• go_buildpack_name: go_buildpack
• python_buildpack_name: python_buildpack
• php_buildpack_name: php_buildpack
• binary_buildpack_name: binary_buildpack
• hwc_buildpack_name: hwc_buildpack

The route_services test group pushes applications which must be able to reach the load balancer of your Cloud Foundry deployment. This requires configuring application security groups to support this. Your deployment manifest should include the following data if you are running the route_services group:

...
properties:
  ...
  cc:
    ...
    security_group_definitions:
      - name: load_balancer
        rules:
        - protocol: all
          destination: IP_OF_YOUR_LOAD_BALANCER # (e.g. 10.244.0.34 for a standard deployment of Cloud Foundry on BOSH-Lite)
    default_running_security_groups: ["load_balancer"]

To run tests that exercise container networking and running application security groups, the include_security_groups flags must be true.

The Windows ASG tests require an unallocated IP on the private network used by the CF deployment, set with the unallocated_ip_for_security_group config value. Environments created by bbl on public clouds can use the default value of 10.0.244.255. vSphere and Openstack environments may require a custom IP.

To run tests that exercise the use of credentials to access a private docker registry, the include_private_docker_registry flag must be true, and the following config values must be provided:

• private_docker_registry_image
• private_docker_registry_username
• private_docker_registry_password

These tests assume that the specified private docker image is a private version of the cloudfoundry/diego-docker-app:latest. To upload a private version to your DockerHub account, first create a private repository on DockerHub and log in to docker on the command line. Then run the following commands:

docker pull cloudfoundry/diego-docker-app:latest
docker tag cloudfoundry/diego-docker-app:latest <your-private-repo>:<some-tag>
docker push <your-private-repo>:<some-tag>

The value for the private_docker_registry_image config value in this case would be ":".

To run tests that involve routing isolation segments, the following config values must be provided:

• isolation_segment_name
• isolation_segment_domain

Read the documentation here for further setup details.

• non-assisted mode means that apps are responsible for resolving Credhub refs for credentials. When the user binds a service to an app, the service broker will store a credential in Credhub and pass the ref back to the Cloud Controller. When the user restages the app, the Cloud Controller will pass the Credhub ref to the app in the VCAP_SERVICES environment variable, at which point the app can make a request directly to Credhub to resolve the ref and obtain the credential. This mode is enabled when cc.credential_references.interpolate_service_bindings is false -- which is the non-default configuration.
• assisted mode means that the Credhub ref will be resolved before the app starts running. As before, when the user binds a service to an app, the service broker will store a credential in Credhub and pass the ref back to the Cloud Controller. This time, when the user restages the app, the Cloud Controller will pass the Credhub ref to the Diego runtime, at which point the launcher (from the buildpackapplifecycle or the dockerapplifecycle components) will resolve the Credhub ref, and store the credential in VCAP_SERVICES for the app to consume. This mode is enabled when cc.credential_references.interpolate_service_bindings is true -- which is the default configuration.

When a test fails, look for the test group name ([services] in the example below) in the test output:

• Failure in Spec Setup (BeforeEach) [34.662 seconds]
[services] Service Instance Lifecycle

If you set a value for artifacts_directory in your $CONFIG file, then you will be able to capture cf trace output from failed test runs, this output may be useful in cases where the normal test output is not enough to debug an issue. The cf trace output for the tests in these specs will be found in CF-TRACE-Applications-*.txt in the artifacts_directory.

To execute all test groups, run the following from the root directory of cf-acceptance-tests:

./bin/test

It's possible execute all test groups, and have tests run in parallel across multiple processes. This parallelization can significantly decrease CATs runtime.

However, be careful with this number, as it's effectively "how many apps to push at once", as nearly every example pushes an app.

In order to run 12 processes in parallel run the following:

./bin/test -nodes=12

For reference here's how many parallel processes the Release Integration team runs with:

If you are already familiar with CATs you probably know that there are many test groups. You may not wish to run all the tests in all contexts, and sometimes you may want to focus individual test groups to pinpoint a failure. To execute a specific group of acceptance tests, e.g. routing/, edit your integration_config.json file and set all include_* values to false except for include_routing then run the following:

./bin/test

To execute tests in a single file use an FDescribe block around the tests in that file:

var _ = AppsDescribe("Apps", func() {
  FDescribe("Focused tests", func() { // Add this line here
  // ... rest of file
  }) // Close here
})

The test group names correspond to directory names.

To see verbose output from ginkgo, use the -v flag.

./bin/test -v

You can of course combine the -v flag with the -nodes=N flag.

This repository uses go mod to manage go dependencies.

All go dependencies required by CATs are vendored in the vendor directory.

When making changes to the test suite that bring in additional go packages, you should use the following workflow:

If you can use the most recent version of a dependency use go mod tidy, otherwise use go get <dependency>@<version>. Both of these require go modules to be enabled via the envrc. Finally use go mod vendor to add the dependencies to the vendor directory.

For tools and assets, please use the helpers/assets/tools.go via the go mod tool workflow

For additional information refer to the official wiki and the official examples repo

Although the default branch for this repository is master, we ask that all pull requests be made against the develop branch. Please run the unit tests and make sure they are passing before submitting. Use ./bin/run_units to run these unit tests.

Note: it is necessary to run the tests from the root of the repo.

There are a number of conventions we recommend developers of CF acceptance tests adopt:

1. When pushing an app:

• set the memory requirement, and use the suite's DEFAULT_MEMORY_LIMIT (DEFAULT_WINDOWS_MEMORY_LIMIT for tests in the windows directory) unless the test specifically needs to test a different value,
• set the buildpack unless the test specifically needs to test the case where a buildpack is unspecified, and use one of config.RubyBuildpack, config.JavaBuildpack, etc. unless the test specifically needs to use a buildpack name or URL specific to the test,
• set the domain, and use the Config.AppsDomain unless the test specifically needs to test a different app domain.

For example:

Expect(cf.Cf("push", appName,
    "-b", buildpackName,                  // specify buildpack
    "-m", DEFAULT_MEMORY_LIMIT,           // specify memory limit
    "-d", Config.AppsDomain,              // specify app domain
).Wait(Config.CfPushTimeoutDuration())).To(Exit(0))

1. Delete all resources that are created, e.g. apps, routes, quotas, etc. This is in order to leave the system in the same state it was found in. For example, to delete apps and their associated routes:

   	Expect(cf.Cf("delete", myAppName, "-f", "-r").Wait()).To(Exit(0))
   

2. Specifically for apps, before tearing them down, print the app guid and recent application logs. There is a helper method AppReport provided in the app_helpers package for this purpose.

   AfterEach(func() {
     app_helpers.AppReport(appName)
   })

3. Document the purpose of your test groups in this repo's README.md. This is especially important when changing the explicit behavior of existing test groups or adding new test groups.

4. Document all changes to the config object in this repo's README.md.

5. If you add a test that requires a new minimum cf CLI version, update the minCliVersion in cats_suite_test.go .